Automation ROI: how to calculate the return on an AI workflow
The return on an AI workflow depends on time saved, adoption, output quality, maintenance and the cost of errors avoided. This article gives owners and operations leads at SMEs a calculation that fits in a spreadsheet, with a worked example.
The simplest automation case starts with time. If a workflow takes ten hours per week and AI can reliably remove five, the value is easy to understand. But time saved is only one part of the business case. A workflow that saves time and introduces rework, risk or low adoption will not pay back.
The calculation below combines four inputs: visible savings, quality gains, adoption and running costs. Its output is a net monthly benefit and a payback period.
Calculate the visible savings
Start with four figures: how often the task happens, the average time per task, the number of people involved and their loaded hourly cost. Loaded cost means salary plus social charges and overheads, which is what an hour of that person's time costs the company.
Then estimate the share of the work the automation can remove after human review. Someone still reads, corrects and approves each draft, so the saving is the old time per task minus the new review time. Take both figures from timed cases where you can. A workflow audit produces them, and the article on designing human review covers the review step itself. Conservative estimates are more useful than optimistic demos.
Add quality and speed
Some AI workflows are valuable because they reduce errors, standardise outputs or shorten response times. A support triage workflow may not remove a role, but it can improve first response time and reduce missed escalations.
Put a figure on these gains only where you can defend it. Rework is the easiest to cost: count how often output comes back for correction today, multiply by the time each correction takes, and estimate how much of that the workflow would avoid. Where a gain cannot be costed, such as a faster first response, list it beside the calculation and leave it out of the total.
Adjust for adoption
A workflow used by half the team returns half the expected value. Multiply the saving by the share of cases you expect to go through the new workflow, and set that share cautiously for the first months.
Expect lower adoption when staff have to open a separate tool and paste text across, and higher adoption when the draft appears in the inbox, CRM or ticket queue where they already work. After launch, replace the assumption with a measurement.
Include maintenance costs
Prompts, integrations, source documents and business rules change. Every automation needs someone to monitor output quality and update the workflow. Include that work in its running cost.
Keep two groups of costs apart. One-off costs are setup and training. Running costs recur every month: subscriptions or usage fees, the owner's time for checks and updates, and support for users. If a provider maintains the system, as in Block Consult's ongoing AI care, maintenance and smaller changes fall within an agreed monthly scope and budget, which enters the calculation as a known figure. Larger developments are costed separately.
The same reasoning applies before commissioning a custom tool. The expected savings have to justify maintenance beyond the first release, as our guide to buying or building AI tools explains.
The automation ROI calculation step by step
- Count the tasks per month and the average time each one takes today.
- Estimate the time each task takes with the automation, including review and correction. The difference is the time saved per task.
- Multiply tasks by time saved and by the loaded hourly cost. This is the gross monthly saving.
- Add the monthly value of errors or rework avoided, where you can put a cost on it.
- Multiply by the adoption rate, the share of cases that go through the workflow.
- Subtract the monthly running costs for subscriptions, maintenance and support. The result is the net monthly benefit.
- Divide the one-off costs for setup and training by the net monthly benefit. The result is the payback period in months.
A worked example with round numbers
Suppose a support team answers 300 customer enquiries a month. The figures are invented for illustration, and each one is an assumption you would replace with your own.
- Volume: 300 enquiries a month.
- Time per enquiry today: 15 minutes.
- Time with an AI draft: 5 minutes to review and correct, so 10 minutes saved.
- Loaded hourly cost: CHF 80.
- Adoption: 75 per cent of enquiries go through the new workflow.
- Running costs: CHF 1,000 a month for subscriptions, usage and the owner's maintenance time.
- One-off costs: CHF 12,000 for setup and training.
The gross saving is 300 enquiries times 10 minutes, which is 50 hours a month or CHF 4,000. At 75 per cent adoption that becomes CHF 3,000. After running costs the net benefit is CHF 2,000 a month, so the one-off costs are paid back after six months. After twelve months the net benefit adds up to CHF 24,000 against CHF 12,000 of one-off costs.
Now change one assumption. At 40 per cent adoption the adjusted saving is CHF 1,600, the net benefit is CHF 600 a month and payback takes 20 months. The tool is the same in both cases. Only the share of enquiries that go through it differs.
Test the estimate before committing
The calculation is a forecast until cases have gone through the workflow in daily work. A pilot with one team replaces the two weakest assumptions, time per task with review and adoption, with measured figures. Our article on taking an AI pilot into daily use covers the evidence a pilot should produce.
In Block Consult's AI readiness and workflow audit, the economics of each candidate workflow are examined alongside its data, systems and people, and estimates are recorded with their assumptions and missing evidence. That makes it possible to compare candidates before choosing a first AI workflow.
Frequently asked questions
- What payback period should SMEs expect for AI automation?
- Focused workflows can pay back in months, but timing depends on setup cost, adoption, maintenance and the value of the work being improved.
- How do you calculate the ROI of an AI workflow?
- Multiply the hours saved per month by the loaded hourly cost, add the value of rework avoided, adjust for adoption and subtract monthly running costs. Dividing the one-off setup and training costs by that net monthly benefit gives the payback period in months.
- Should we automate a broken process?
- Usually no. First simplify the workflow, remove unnecessary steps and clarify ownership. AI works better on a clean process.
- How do we measure adoption?
- Track how often the workflow is used, which teams use it, where users override outputs and whether the automation changes the target metric.
Related reading
Let's build something durable.
Tell us about your project. We reply within two business days.

